Gandal DGE: me vs article

load('./../Data/Gandal_RNASeq.RData')

DE_info_paper = read.csv('./../Data/Gandal_paper_DGE.csv')[,1:7] %>% mutate('ID'=X)

DE_info_all = DE_info %>% left_join(DE_info_paper, by='ID') %>% 
              filter(complete.cases(.)) %>%
              left_join(SFARI_genes, by='ID') %>%
              mutate(`gene-score` = ifelse(is.na(`gene-score`), 
                                           ifelse(ID %in% GO_neuronal$ID, 'Neuronal', 'Non-Neuronal'), 
                                           `gene-score`))

Log fold change comparison

The signs are in the opposite direction (took CTL instead of ASD as 1? maybe it was me … I think not, but could have been)
92% of the genes are classified the same in both analysis

ggplotly(DE_info_all %>% mutate('logFC_me' = logFC, 'logFC_Gandal'=All.logFC) %>% 
         ggplot(aes(logFC_me, logFC_Gandal)) + 
         geom_point(alpha=0.3, aes(id=ID, fill=`gene-score`, color=`gene-score`)) +
         scale_color_manual(values=gg_colour_hue(9)) + scale_fill_manual(values=gg_colour_hue(9)) +  
         ggtitle(paste0('Corr=',round(cor(DE_info_all$logFC, DE_info_all$All.logFC), 2))) +
         geom_abline(slope=-1, color='#808080', size=0.5) + theme_minimal() + coord_fixed() + 
         geom_hline(yintercept=log2(1.5), color='#a6a6a6') + geom_hline(yintercept=-log2(1.5), color='#a6a6a6') +
         geom_vline(xintercept=log2(1.5), color='#a6a6a6') + geom_vline(xintercept=-log2(1.5), color='#a6a6a6'))

round(table(abs(DE_info_all$logFC)>log2(1.5), abs(DE_info_all$All.logFC)>log2(1.5))/nrow(DE_info_all)*100,1)

##        
##         FALSE TRUE
##   FALSE  89.6  0.6
##   TRUE    6.8  3.1

Adjusted p-value comparison

My adjusted p-value seems to be a lot stricter, could be because I took a larger amount of genes (33K vs 20K)
76.4% of the genes are classified the same way in both experiments

ggplotly(DE_info_all %>% mutate('adj.P.value_me' = adj.P.Val, 'adj.P.value_Gandal'=All.adj.P.Val) %>% 
         ggplot(aes(adj.P.value_me, adj.P.value_Gandal)) + 
         geom_point(alpha=0.1, aes(id=ID, fill=`gene-score`, color=`gene-score`)) +
         scale_color_manual(values=gg_colour_hue(9)) + scale_fill_manual(values=gg_colour_hue(9)) +  
         ggtitle(paste0('Corr=',round(cor(DE_info_all$adj.P.Val, DE_info_all$All.adj.P.Va), 2))) +
         geom_abline(color='#808080', size=0.5) + geom_hline(yintercept=0.05, color='#a6a6a6') + 
         geom_vline(xintercept=0.05, color='#a6a6a6') + theme_minimal())

round(table(DE_info_all$adj.P.Val<0.05, DE_info_all$All.adj.P.Val<0.05)/nrow(DE_info_all)*100,1)

##        
##         FALSE TRUE
##   FALSE  68.0  1.1
##   TRUE   22.5  8.4

Significant genes

75.9% of the genes have the same classification in both analysis
Gandal classifies

DE_info_all = DE_info_all %>% mutate('significant_me' = abs(logFC)>log2(1.5) & adj.P.Val<0.05,
                                     'significant_Gandal' = abs(All.logFC)<log2(1.5) & adj.P.Val<0.05)

round(table(DE_info_all$significant_me, DE_info_all$significant_Gandal)/nrow(DE_info_all)*100, 1)

##        
##         FALSE TRUE
##   FALSE  69.3 21.1
##   TRUE    3.0  6.6